Ulcerative colitis AI identifies activity vs. remission, predicts future flareups
Researchers across the pond have developed and externally validated an AI model that can predict flareups of ulcerative colitis.
The computer-aided diagnosis (CAD) system largely works by ascertaining disease activation or remission using three separate endoscopy scoring systems normally administered by expert pathologists or gastroenterologists.
The work is described in a study published online March 3 in Gastroenterology [1]. Corresponding and co-lead author is Marietta Iacucci MD, PhD, of University College Cork in Ireland and the University of Birmingham in England. The senior author is Vincenzo Villanacci, MD, of the University of Brescia in Italy.
The team began the project by grading 535 biopsies from 273 patients according to the PICaSSO Histologic Remission Index (PHRI), Robarts Histopathology Index (RHI) and Nancy Histological Index (NHI).
They used these scores to train a convolutional neural network classifier how to distinguish colitis remission from activity on a subset of 118 of the 535 biopsies before calibrating the AI with 42 biopsies and pretesting it on 375.
The team further tested the model to predict flareups 12 months out from colonoscopy.
For ground truth, Iacucci and colleagues used biopsy assessments from a panel of experienced pathologists, six of whom had particular expertise in irritable bowel disease (IBD).
They found their AI model distinguished histological activity/remission with sensitivity and specificity of 89% and 85% on the PHRI scoring system, 94% and 76% (RHI) and 89% and 79% (NHI).
The model predicted endoscopy-visible remission vs. activity with accuracy of 79% to 82%.
Both histology and outcome prediction were confirmed in an external validation cohort of 154 biopsies from 58 patients.
In their discussion, the authors suggest their model can “expedite, standardize and enhance” histological assessments of ulcerative colitis patients in daily medical practice as well as in future clinical trials.
They call for additional research to include colorectal dysplasia detection and to “combine histologic and endoscopic AI models into an integrated tool to further improve disease monitoring and prediction.”
The study is available in full for free (PDF).
Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.